Re: regarding build speed on riscv64

On Tue, Sep 01, 2026 at 10:18:28AM -0400, Dominique Belhachemi wrote:
> On Tue, Sep 1, 2026 at 6:50 AM Adrian Bunk <[email protected]> wrote:
> 
> > On Mon, Aug 31, 2026 at 06:39:07PM -0400, Dominique Belhachemi wrote:
> > > On Mon, Aug 31, 2026 at 4:55 PM Aurelien Jarno <[email protected]>
> > wrote:
> > >  ...
> > >
> > > In parallel, we continue to follow upstream kernel support for the K3,
> > > > with the hope we can get everything in 7.4. Currently the PCI
> > controller
> > > > [2] and the thermal management chip [3] are missing, and a few smaller
> > > > fixes that will likely go through the stable trees anyway. The bad news
> > > > is that it 7.4 is currently scheduled for January [4], so 7.3 will be
> > > > the LTS version that will go into Forky. This means we'll need to work
> > > > with the kernel team to backport the corresponding patches into 7.3.
> > > >
> > > >
> > > Would it be possible for Debian to procure and distribute 10 to 20 K3
> > > boards to interested DDs?
> >
> > Money in Debian is a mess, and your suggestion is somewhere around € 10k.
> >
> Why is it a mess? What happened?

Have you ever seen an annual financial report or even a budget for Debian?

> > Having hardware available to individual maintainers for local testing and
> > debugging would certainly help unblock and strengthen the riscv64 port.
> >
> > What problems do you plan to debug where you need physical hardware?
> >
> Boot issues, kernel panics, timeouts, extensions, GPU bringup, 3D graphics, AI.

That's not work for "individual maintainers", that's porter work.

It's appreciated if you want to participate in porting, but lack of
hardware is not biggest the obstacle for that.

Topics like GPU or AI are also not relevant for this discussion, which 
is about the buildds.

All 8 release architectures combined do not have 10 DD porters,
sending 10-20 pieces of hardware to interested DDs wouldn't
magically create 10-20 DDs with the skills, interest and
available time to be active porters.

A Raspberry Pi costs nearly nothing and many DDs have at least one,
but the number of constantly active arm porters is close to zero.

20-30 years ago porting Debian to architectures like m68k or hppa was
a cool hobby for computer science students. Students today are more 
interested in other topics like AI.

Some documentation still requires a port to have at least 3 or 5 active porters.
Just a dozen years ago even architectures like hurd or kfreebsd had no 
problem reaching that number.
Requiring 5 active porters for a release architecture today would imply 
that we stop doing releases, due to not having a single release architecture.

> > Like every release architecture (and many ports architectures) riscv64
> > has a porterbox, which is usually what individual maintainers need for
> > debugging issues.
> >
> The porterbox has limited value to us:
> isa : rv64imafdc_zicntr_zicsr_zifencei_zihpm_zca_zcd

Who is "us"?

For the purpose of debugging FTBFS issues on the buildds, there is 
nothing better than a machine that has exactly the same hardware
as the buildds.

It needs skill and time to debug and fix bugs like #1145420,
even with a K3 porterbox available today there wouldn't be
a queue of people willing and able to debug such bugs.

> Best
> -Dominique

cu
Adrian